ANTIGROUPWARE AND SECOND MESSENGER
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
AntiGroupWare and Second Messenger M I Norton, J M DiMicco, R Caneel and D Ariely Decision-making in groups has great potential due to the possibilities for pooling ideas and sharing knowledge, but also great drawbacks due to the social pressures inherent in these situations that can limit free exchange of these ideas and knowledge. This paper presents two technology-based approaches to improving group decision-making, Second Messenger and AntiGroupWare. Second Messenger — a system that encourages groups to change their interaction styles during meetings — is designed to improve meetings, while AntiGroupWare — an on-line polling system that allows companies to gather information through flexible, iterative polling of its employees — is designed to avoid them altogether. 1. Introduction This is just one example of where group deliberation can result It was the morning of March 19, 2003 in the White in poor decision-making. In one of the most infamous episodes House Situation Room, just hours before President in American foreign policy, John F Kennedy and his advisors, Bush’s 48-hour ultimatum to Saddam Hussein to get after extensive group discussion of the plan, signed off on the out of Iraq was to expire. President Bush had just polled ‘Bay of Pigs’ invasion of Cuba in 1961, a plan that had virtually his war council for any last-minute reservations about no chance of success [2]. the war plan. Hearing none, he issued the ‘execute’ command to General Tommy R Franks. sadly, the intuitions about The New York Times, April 14th 2003 the benefits of group Organisations are called upon to make decisions almost decision-making seem to be continuously, and one of the most common models for doing so utilises a format with which most are all too familiar — overly optimistic gathering relevant individuals together in a meeting to discuss the issues at hand and reach a decision to be implemented. In one of the most infamous cases of American corporate This approach seems at first glance to be an ideal strategy, but chicanery, nearly every board vote taken at the Enron 50 years of research on the drawbacks of making decisions in Corporation was unanimous [3] — despite the fact that groups suggests that this model — despite its widespread use Enron’s policies were disastrous. How can groups of qualified, — can lead to suboptimal decision-making [1]. In the example intelligent individuals make such suboptimal decisions when above, two factors stand out for their potentially adverse gathered together? How is it that pooling the knowledge and impact on decision-making in groups — authority structures expertise of individuals sometimes fails to improve decision- and conformity pressures. As the clear authority, Bush spoke making? Sadly, the intuitions about the benefits of group first, making his position clear, and then polled his decision-making seem to be overly optimistic; groups subordinates for their opinions. Additionally, by asking each sometimes bring the worst out of individuals. person for their opinion one-by-one, the pressure for the last Cabinet member to agree with the plan increased. Thus while This paper presents two systems, Second Messenger and it appears that the decision to attack Iraq was a unanimous AntiGroupWare, designed to remedy these problems of decision made by a group of highly skilled individuals, the decision-making in groups. Second Messenger focuses on pressures on this group to agree with the authority figure and improving decision-making during meetings, while conform to the majority opinion may have undermined its AntiGroupWare focuses on improving decision-making out of ability to make optimal decisions. meetings, allowing meetings to be avoided altogether. In the BT Technology Journal • Vol 22 No 4 • October 2004 83
AntiGroupWare and Second Messenger following sections, we review the literature on group biases, wrong answer [6]. In some cases, up to 75% of people would discuss the potential for technology to improve these biases, give the wrong answer, simply to fit in with their peers. and then describe how Second Messenger and AntiGroupWare Because real life decisions are, if anything, even more might help. ambiguous than the ones Asch used in his experiments, conformity pressures can be powerful in inducing individuals to acquiesce publicly to decisions with which they disagree 2. Two key problems with groups — privately. authority and conformity Two factors particularly likely to have an adverse impact on In summary, the presence of authority structures and decision-making in meetings are authority structures and conformity pressures can lead to less information-sharing and conformity pressures. Those in positions of authority or less discussion of conflicting viewpoints, resulting in influence may or may not have greater expertise or knowledge suboptimal decisions. This is not to say that all social influence than others, but due to the tendency of individuals to defer to in groups will have negative consequences, of course — it is authority figures (following their advice and endorsing their sometimes the case that those in authority are present recommendations even when these decisions are flawed or because they may be better qualified to guide discussion and even harmful), those in positions of power can exert inordinate shape policy, or that consensus is dictated not by conformity influence over the tone of the discussion and the eventual but by the facts. Rather, what the present paper seeks to outcome. A classical illustration of the power of authority was address are those cases where the quality of the group provided in a set of famous experiments by Stanley Milgram, decision is adversely impacted by social pressures of which the where individuals were willing to administer what they thought group is unaware and in response to which it may be acting were harmful, or even fatal, electric shocks to other blindly, and thus succumbing to the negative effects of group participants simply because a ‘scientist’ in a white lab coat interaction. told them to do so [4]. In the real-world setting of airplane flight crews, the term ‘captainitis’ was coined to describe the many near-disaster airplane incidents in which a member of 3. Technology and group decision-making the flight crew noticed a problem but did not to mention it to The underlying assumption behind both Second Messenger the captain, presumably because they did not trust their own and AntiGroupWare is that technology can be used to improve judgement over the captain’s [5]. Though non-authority group decision-making, a concept that has been increasingly figures often have information and expertise that is relevant to studied over the last twenty years. When technology is the decision at hand, concerns regarding authority can introduced as a communications medium between group prevent individuals from airing their views publicly. To the members, the way we communicate fundamentally changes, extent that it inhibits this disclosure of information, the and in some cases this can create additional group process presence of authority figures can have a seriously adverse problems. For example, videoconferencing is commonly used impact upon group decision-making. for meetings within companies, yet studies have shown that there is less trust between individuals and that it is more difficult to establish common ground over video than in face- the presence of authority to-face interactions [7—9]. When communication is text- based, such as with chat, e-mail or, increasingly, instant- structures and conformity messenger, decision-making tasks take more time and produce lower rates of task accuracy [10—12]. pressures can lead to less information-sharing and less Text-based applications also cause individuals to make more declarative position statements and less information-based discussion of conflicting statements [13], making productive information-based meetings more difficult to hold. In the specific areas of group viewpoints polarisation and information sharing, which relate back to our issues of authority and conformity, it has been shown that groups can experience more group polarisation [10, 12] and Even in the absence of authority figures, the mere presence of less information sharing [14] when communicating in a text- others can have unexpected and negative effects on group based environment. decision-making. As adolescents know all too well, the opinions, preferences, and behaviour of one’s peers can have At the same time, some research shows that technology — an inordinate impact on personal choices. When several when used appropriately — can help improve group individuals express one opinion in a meeting, other individuals performance, by minimising inhibitory social pressures [15]. In can conform to this emerging group norm, even when that systems that allow groups to anonymously interact and opinion is objectively wrong. Solomon Asch demonstrated this contribute to a brainstorming task, studies have shown tendency in a set of classic experiments in which participants marked improvement in systems that are anonymous over heard a series of other people give what was clearly a wrong those that identify the users [16, 17]. Along these lines, we answer to a simple judgement task. Asch’s results revealed outline below two systems, Second Messenger and that the more people who had given the wrong answer, the AntiGroupWare, that attempt to utilise the positive aspects of more likely the study participant — the only person who was technology for the purpose of improving group decision- not in league with the experimenter — was to also give the making. 84 BT Technology Journal • Vol 22 No 4 • October 2004
AntiGroupWare and Second Messenger 4. In-meeting decisions — Second removed the textual content of Second Messenger and instead Messenger displayed the quantity of spoken comments made by each Second Messenger is a system that focuses on encouraging a participant. As a group interacts, the display dynamically group to change its interaction during a meeting, based on adjusts the bars of a histogram to indicate the relative the idea that making a group aware of its potentially flawed participation rates (see Fig 2 for a screenshot). The system is interaction while it is taking place affords the group the built again with a client/server architecture, where each client opportunity to correct its behaviour. When interacting face-to- machine determines when someone is speaking by detecting face, individuals utilise their natural strengths in the sound level from individual microphones. communicating opinions and intentions, yet also naturally overly rely on the group’s dominating opinions, as already discussed. We postulate that a system observing the natural dynamics of a group interaction can detect skewed group processes and can then influence the group, through visual displays of social information, to alter its interaction to focus more successfully on the breadth of ideas in a discussion. over over As an initial proof-of-concept and to explore the possibilities participating for the type of system proposed, the Second Messenger participating application [18] augments a face-to-face interaction by under providing a real-time visual summary of the verbal comments under made during a group meeting. Second Messenger is built around a client-server architecture where each group member wears a microphone that sends his/her spoken dialogue to a Fig 2 Histogram display of speaker participation. client running IBM’s ViaVoice, using a trained voice model to recognise individual speech. When the user speaks, the The behavioural experiment’s protocol was based on an transcribed words are sent to the server for analysis and information-sharing task previously used by Hollingshead [14] filtering. After filtering for accuracy and uniqueness, the text in which each subject group was given a task to solve, the phrases appear on the shared display and can be manipulated successful completion of which depended on the amount of and organised using a mouse, making it a tool for organising information shared by the group. The study’s control the group discussion (see Fig 1 for a screenshot). In addition to condition had groups complete two tasks with no being a tool though, Second Messenger attempts to increase technological assistance; the experimental condition had the the amount of discussion around diverse viewpoints by display projected on to the wall during the second of two increasing the visibility of the comments made by group tasks. members who speak less frequently and by filtering out the comments of group members who verbally dominate. The main finding from the study was that the over- participators 1 decreased the amount they spoke in the The next step of this research was to run a controlled experimental condition significantly more so than those in the behavioural experiment to understand how a simple display of control condition (p
AntiGroupWare and Second Messenger 15 task 1 individuals to rank multiple options, so organisations can assess support for a fuller set of alternatives (for example it is 10 possible that all individuals differ on their most preferred option but they all agree on the second best option or the 5 least preferred option). An example of such a method is known over-participators as the Borda preferendum, in which ranks are associated with participation rate control experimental graduated points (e.g. 1st ranked options gets 15 points, 2nd 0 ranked gets 10, 3rd gets 8, etc), a system that has been used control experimental effectively in other domains (e.g. professional sports ‘Most -5 under-participators Valuable Player’ awards). -10 5.2 Iterative voting -15 Another feature of the system is that polls can allow for iterative voting. An advantage of this is that individuals can -20 task 2 dynamically react to the information available in others’ responses. This approach is in contrast to the way voting is usually carried out, which can be characterised as a ‘one-shot Fig 3 Changes in participation rates by participation category simultaneous process’ much like sealed-bid auctions. This and condition. function bears some resemblance to the Delphi technique developed at the Rand Corporation [20], a multiple-round does have an impact on the behaviour of a group. Our planned process by which managers can pool the opinions of selected future work is to continue to build visual displays of ongoing individuals. These procedures were often time-consuming, social interaction and develop a deeper understanding of how involving multiple rounds of paper questionnaires, an issue these displays encourage and discourage certain behaviours. avoided by using on-line mechanisms. The nominal group Our goal is to produce a set of interfaces that increase the technique [21], in which participants meet in groups but first diversity of a group’s discussion and assist in improving the write their opinions about issues privately before sharing with quality of the interaction. other group members, is another system designed to improve group-decision-making; again, however, while the technique has benefits, it requires lengthy meetings, and a moderator. 5. Out-of-meeting decisions — The ability to express opinions iteratively — with little cost and time — makes AntiGroupWare resemble an information- AntiGroupWare sorting mechanism such as the stock market or an open AntiGroupWare is an alternative approach to improving auction. To illustrate this point, consider an open auction, decision-making that enables groups to avoid meeting where the purpose is price discovery and there are benefits in altogether. The AntiGroupWare application is an on-line letting some people voice their opinions (prices) early, letting polling system that companies can use to make decisions others observe these opinions and then considering this new through flexible, iterative polling of its employees. Thus the information together with their prior opinions. In particular, system is designed to tap into organisational knowledge, this process can help in information discovery in cases where which has received increasing attention as a useful source of people can get additional insights from the opinions of others. information for corporations. Aggregation of estimates made When bidding on an antique divan, for example, the fact that by Hewlett-Packard employees were shown to be a better others value it highly decreases the probability that the divan predictor of the value of HP stock than the predictions of, for is a fake. Thus the means by which the opinions of others are example, the highly trained analysts hired by HP, and the revealed can be beneficial to decision makers. We propose ‘average Americans’ who participate in the Iowa Electronic that similar systems can be useful for voting and opinion Markets have proved to be more accurate than experts in mechanisms more generally, in those cases where the forecasting the outcomes of US presidential campaigns. To information contained in the opinions of others can inform harness the strength of collective knowledge and to improve one’s own understanding. AntiGroupWare has the facility to upon existing models, AntiGroupWare has three key features allow dynamic voting that lasts over multiple stages where — flexible voting, iterative voting, and proxy voting. individuals can revisit their votes, see where the winds are blowing in the overall decision, and if they so desire change their opinions. This process, while more involved than one- 5.1 Flexible voting shot voting, clearly has the potential to improve decision When polling is limited to basic voting procedures, such as quality. allowing single ‘Yes/No’ responses to single issues, the effectiveness of the poll is limited. AntiGroupWare provides a much-increased level of flexibility in collecting responses. 5.3 Proxy voting Specifically, the AntiGroupWare system allows individuals to A final feature of AntiGroupWare is its ability for individuals to rate multiple options on multiple dimensions. For example, a assign their votes to other individuals who may be more proposal can be rated for its maximum benefit and its risk. As knowledgeable, an opportunity usually restricted to a result, the separate evaluation of these two estimates offer a management. First, the system allows managers to assign fuller sense of the expected payoff of the proposal than a greater weight to the votes of those individuals deemed to be simple ‘Yes/No’ would have. AntiGroupWare also allows most qualified. More interestingly, AntiGroupWare also allows 86 BT Technology Journal • Vol 22 No 4 • October 2004
AntiGroupWare and Second Messenger individuals themselves to assign their votes (proxy-voting) to 6. Conclusions someone of their choosing, if they feel that another person Alfred Sloan, who ran General Motors from 1923 to 1956, would be better equipped to make a decision in some domain. once said during a meeting: ‘Gentlemen, I take it that we are For example, experts in recycling, who are recognised by their all in complete agreement on the decision here. Then, I peers as such, might get assigned voting rights by their peers propose that we postpone further discussion... to give for topics related to recycling. Importantly, in order to ourselves time to develop disagreement and perhaps gain maintain anonymity, those who have been assigned additional some understanding of what the decision is all about’ [3]. weight make their choices or judgements without the Alfred Sloan was not a management genius for nothing. Years knowledge of their additional power of influence — or the later we are following in his footsteps with an increased source(s) of that increased influence. This system thus has the understanding of the reasons underlying poor group decision- potential to tap into organisational knowledge not captured making, with more advanced approaches for solving them. on organisational flow charts (and thus not necessarily The technology-based approaches to group decision-making available to managers), and better utilises organizational presented here, Second Messenger and AntiGroupWare, can knowledge by allocating decision-making authority to the minimise group decision biases while at the same time best-qualified individuals. improving data aggregation more generally both in and out of meetings. Second Messenger allows for better sharing of It will not always be the case that individuals know who is most information, and thus more optimal decisions, while knowledgeable, or that they will assign their votes in the most AntiGroupWare is designed to improve and streamline optimal manner, yet results such as those at Hewlett Packard decision-making, at the same time sparing people lengthy, and in the Iowa Electronic Markets suggest that this kind of and often unproductive meetings. knowledge can be extremely useful. Acknowledgements 5.4 Summary The anonymity inherent in AntiGroupWare addresses the The work described in this paper has been supported by the negative impact of both authority and conformity — the Digital Life and information:organized consortia at the Media former by ensuring that management cannot monitor which Lab. Additional thanks go to the extremely helpful anonymous employee has voted for which proposal, and the latter by reviewers. allowing for votes to be taken in private where people are licensed to express their true opinions. The iterative voting functionality allows for some social influence to occur (as References people may change their votes to match a growing consensus) 1 Davis J H: ‘Some compelling intuitions about group consensus but because the voting is anonymous the social influence decisions, theoretical and empirical research, and interpersonal aggregation phenomena: Selected examples, 1950—1990’, caused by iterative voting may be more informational in Organizational Behavior and Human Decision Processes, 52, nature — as people gather information about others’ pp 3—38 (1992). opinions, they may come to change their opinion not due to 2 Janis I: ‘Groupthink’, Boston, MA, Houghton Mifflin (1982). social pressure but due to new information revealed through the voting process. 3 Surowiecki J: ‘Board Stiffs’, The New Yorker (8 March 2004). 4 Milgram S: ‘Obedience to authority: An experimental view’, New Finally, the proxy voting functionality addresses issues of York, Harper and Row (1974). authority by standing authority on its head, and allowing 5 Foushee M C: ‘Dyads and triads at 35,000 feet: Factors affecting employees to designate authority as they choose, rather than group process and aircraft performance’, American Psychologist, being bound by institutional authority structures. 39, pp 885—893 (1984). 6 Asch S E: ‘Studies of independence and conformity: I. A minority ‘... give ourselves time to of one against a unanimous majority’, Psychological Monographs, 70, No 416, p 70 (1956). develop disagreement and 7 Isaacs E and Tang J C: ‘What video can and cannot do for collaboration’, Multimedia Systems, 2, pp 63—73 (1994). perhaps gain some 8 O'Malley C, Langton S, Anderson A, Doherty-Sneddon G and understanding of what the Bruce V: ‘Comparison of face-to-face and video mediated interaction’, Interacting with Computers, 8, pp 177—192 (1996). decision is all about ...’ 9 Rocco E: ‘Trust breaks down in electronic contexts but can be repaired by some initial face-to-face contact’, Conference on Alfred Sloan Human Factors in Computing Systems (CHI98), Los Angeles (1998). 10 Kiesler S and Sproull L: ‘Group decision making and The AntiGroupWare application is currently operational with communication technology’, Organizational Behavior and Human two of the three features outlined above — flexible and Decision Processes, 52, pp 96—123 (1992). iterative voting — and we are planning to test the system at 11 Ochsman R B and Chapanis A: ‘The effects of 10 communication Ford Motor Company’s new product development sector, to modes on the behavior of teams during co-operative problem- assist Ford in gaining better feedback from employees about solving’, International Journal of Man-Machine Studies, 6, new models and options for product lines. pp 579—619 (1974). BT Technology Journal • Vol 22 No 4 • October 2004 87
AntiGroupWare and Second Messenger 12 Siegel J, Dubrovsky V, Kiesler S and McGuire T W: ‘Group processes in computer-mediated communication’, Organizational Joan Morris DiMicco is a PhD student in Behavior and Human Decision Processes, 37, pp 157—187 the Electronic Publishing group at the (1986). Media Lab. She holds a BS in applied mathematics from Brown University and 13 Reid F, Ball L et al: ‘Styles of group discussion in computer- an MS in media arts and sciences from mediated decision making’, British Journal of Social Psychology, MIT. Her master’s research focused on 36, pp 241—262 (1997). dynamic pricing in agent-driven market- places. Prior to the Media Lab, she worked 14 Hollingshead A B: ‘Information suppression and status persistence in the ‘dot.com’ domain, introducing the in group decision-making — the effects of communication media’, Internet’s first secure payment system, Human Computer Research, 23, pp 193—219 (1996). created at First Virtual Inc, and next at 15 Dennis A R, George J F, Jessup L M, Nunamaker J F and Vogel D R: Open Sesame, where she built Web sites ‘Information technology to support electronic meetings’, MIS that learned users’ preferences by Quarterly, 12, pp 591—624 (1998). observing their on-line behaviour. Her doctoral research continues to explore how technology can observe 16 Connolly T, Jessup L M and Valacich J S: ‘Effects of anonymity and and recognise patterns in human behaviour, and in particular how this evaluative tone on idea generation in computer-mediated groups’, information can be used to assist groups in improving their interaction Management Science, 36, pp 97—120 (1990). and decision-making processes in face-to-face settings. 17 Jessup L M, Connolly T and Galegher J: ‘The effects of anonymity on GDSS group process with an idea-generating task’, MIS Quarterly, 14, pp 313—321 (1990). Ron Caneel is a master’s student in MIT’s 18 DiMicco J M and Bender W: ‘Second Messenger: Increasing the program in media arts and sciences. Visibility of Minority Viewpoints with a Face-to-face Collaboration Tool’, Proceedings of the ACM Conference on Intelligent User He received an MS in computer science Interfaces (2004). from the Swiss Federal Institute of Technology (ETH) in 2001 with a minor in 19 DiMicco J M, Pandolfo A and Bender W: ‘Influencing Group psychology from the University of Zurich. Participation with a Shared Display’, Proceedings of the ACM Conference on Computer Supported Cooperative Work (CSCW His research interests are group decision- 2004) (in submission). making and modelling and analysing 20 Brown B B: ‘Delphi Process: A methodology used for the elicitation group interactions. of opinions of experts’, Santa Monica, CA, Rand Corporation (1968). 21 Delbecq A L, Van de Ven A H and Gustafson D H: ‘Group techniques for program planning’, Glenview, IL, Scott, Foresman Dan Ariely holds a joint appointment (1975). between MIT’s Program in Media Arts and Sciences and Sloan School of Management, and is the principal Michael I Norton is a post-doctoral fellow investigator of the Lab’s eRationality at MIT’s Sloan School of Management and group. He is interested in issues of the MIT Media Lab. He received his PhD in rationality, irrationality, decision-making, social psychology from Princeton behavioural economics, and consumer University. welfare. Projects include examinations of on-line auction behaviours, personal His research interests include judgement health monitoring, the effects of different and decision-making; the nature and pricing mechanisms, and the development consequences of political correctness, life of systems to overcome day-to-day satisfaction and well-being (in domains irrationality. He received a PhD in ranging from religious practices to marketing from Duke University, an MA and PhD in cognitive consumer behaviour), and political debate psychology from the University of North Carolina at Chapel Hill, and a strategy. BA in psychology from Tel Aviv University. 88 BT Technology Journal • Vol 22 No 4 • October 2004
You can also read